First off, nice job on the thread. I've been wanting to do something like it for a while but getting it out correctly always seemed like too much thought at the time :).
One of the things I love about pokemon is that all of the bluffs, fake-outs, feints, counter-feints, counter-counter feints, etc. come from a deeper source than pokemon strategy. With a general picture of the metagame and an understanding of game mechanics, the choices we find ourselves making may as well not be pokemon related. Here are some examples, which take only a little imagination to apply to real world competition.
"Is the opponent running a standard set, or an oddball one?"
"Is this strange behavior desperation or a bluff? Is now the time to hammer home my victory or to play it safe?"
"Can I get away with doing the same thing twice, or should I do something less predictable?" (this applies to a lot of situations, such as the one given in this post.)
"Will my opponent predict my move, predict my prediction, predict my over-prediction... etc. ?"
"How does what I've seen over the course of this battle help me to answer the previous four questions?"
My point is, the bulk of the decision making involves "human" considerations, not "pokemon" considerations. Even though some of the above questions contained pokemon terminology, they could have been replaced by any other set of rules and operations, and the "hard" part of the question would still be the same.
I made that point because I think that ultimately this is a more focused discussion if we look at it this way. It is certainly possible for a person to be adept at "figuring out" other people, just as it is certainly possible to be "good" at prediction. I think what is really happening is you are applying your understanding of human judgement to the metagame. The real decision making comes from your understanding of your own judgement, projected onto another person, and edited with your experiences in the battle / in general.
However, I largely agree with the poster that probability also plays a big role in circumstances involving prediction. Some situations have a cut and dry best move to use, and some of them really don't. When a prediction situation is dependent on the opponent's move, then you have to weigh your opponent's options as if they were your own to figure out what he will do. However, there are some situations in which the success of both players' moves are so reliant on what the other player will do that neither player can make a rational decision about the outcome.
The most extreme example is something like a +2 atk Toxicroak vs. an offensive Crobat at 50% health. Crobat can outspeed and OHKO, but Toxicroak can Sucker Punch as Crobat attacks, in which case Toxicroak wins. However, Crobat can Roost to 100% health as Toxicroak sucker punches, allowing him to survive a sucker punch and thus winning the game (assume they are last pokemon). However, if Toxicroak uses Earthquake while Crobat Roosts, Toxicroak wins. If the event that both players know each other's sets (very plausible late-game), this situation is primarily a guessing game. You can try to look at how your opponent has played, but honestly the two choices are so obvious and arbitrary that even knowing your opponent like the back of your hand doesn't tell you what they will do.
That was an extreme example, but there are other examples that are similar, but in a more subtle way. For instance, cut Toxicroak's health and replace Crobat with a sub Espeon or Mismagius. Now, sub is a slightly more attractive choice than attacking, because you can do it multiple times, increasing the likelihood that your opponent sucker punches on a sub (probablility apparent already). Now we have somewhat weighted choices, since sub is a safer bet for Mismagius, they are more likely to use sub (at least at first). The Toxicroak user knows this (hopefully) and so Poison Jab becomes the safer choice, too. However, anyone who has been in this situation will know that the Mismagius user won't always use sub - at some point they will choose shadow ball, sometimes sooner sometimes later. This choice bears in it the same element of randomness as the previous, more extreme, example, despite the fact that some choices are safer than others. The players are trying to throw each other, and hoping to pick the right move each time. My point here is that even choices in which some choices are safer than others have an element of chance, based on the random or pseudorandom (depending on your philosophy) aspects of your opponent's decisions.
Let's get away from Sucker Punch now. A similar condition could occur where two stallers are trying to poison each other. Lets say you have a match-up of sub-roost Moltres vs. a stallish Dragonite or Salamence. You each want to catch each other with a toxic, but the Moltres user is going to try and substitute against the toxic, hoping to get a free toxic. The dragon user knows the Moltres can take a dragon claw and still poison them back. However, Moltres is likely to sub against a toxic and so the dragon user might try to whittle down his subs until he has to roost, then use toxic. The moltres user, trying to avoid this, will roost sooner than necessary, easily taking a dragon claw and starting the process again. The dragon user, if he doesn't figure this out the first time will figure it out soon enough, and might try to throw in a toxic randomly to catch the roost, which is also being randomly chosen to throw off the enemy. This may sound complicated, but it's happened to me numerous times, and the level of understanding I described for both players is certainly there. This example is vastly different than the other two mentioned. Not only is it based on stall rather than offense, but Moltres easily has the upper hand. Chance isn't limited to close matches and make-or-break decisions. This is another great example of when prediction is affected by things that have to do with the human element, but not by skill.
Basically, the answer to whether prediction is luck or skill based is yes. Some predictions are completely luck based (or perhaps "guess" based is a better way of saying it), while others have aspects of strategic thinking behind them. However, the same mechanics that make those ridiculous prediction wars guess based also play at least a small role in every prediction we make. At the end of the day, we are relying on what another human being is going to do, and we all know how fickle humans are.
